Bush v Gore: Prolegomenon to an Assessment

Transcription

1 Bush v Gore: Prolegomenon to an Assessment Richard A. Posnert The Supreme Court's decision terminating the Florida recount and, in consequence, effectively confirming George W. Bush as President has been fiercely criticized by liberal critics. One theme of the critics, though directed less at the decision itself than at its effect, is that, had the recount continued, Al Gore would have been shown to be the "real winner" of the Florida presidential vote. That is incorrect. The recount ordered by the Florida Supreme Court might have given Gore a popular-vote majority. But the recount should never have been ordered. There was no basis for it in Florida law and no reason to believe that Gore had "really" won the election. The basis of that belief was a misunderstanding of statistics and of Florida law. A separate question, one I examine elsewhere,' is whether the U.S. Supreme Court should have intervened. Separate, but not unrelated; that the Florida court was acting arbitrarily was the premise of the equal protection argument that the Supreme Court eventually accepted, and that it was acting in violation of Florida law was the premise of the Article II argument that three Justices found persuasive and that I consider the stronger of the two arguments. I. WOULD A PROPERLY CONDUCTED RECOUNT HAVE PRODUCED VICTORY FOR GORE? After the machine recount of the Florida votes and the addition to the tally of the late-arriving absentee ballots, Bush was ahead by only 930 votes out of almost six million cast in Florida. It was natural to suppose that the voting machines might have made enough mistakes (even after the machine recount) to have given Bush his victory, t Judge, U.S. Court of Appeals for the Seventh Circuit; Senior Lecturer in Law, The University of Chicago. I thank Bryan Dayton for his very helpful research assistance, and Christopher DeMuth, John Donohue, Frank Easterbrook, Eldon Eisenach, Einer Elhauge, Elizabeth Garrett, Dennis Hutchinson, Lawrence Lessig, Michael McConnell, Edward Morrison, Eric Posner, Stephen Stigler, David Strauss, and Cass Sunstein for their very helpful comments on earlier versions of this Essay. 1 See Richard A. Posner, Florida 2000: A Legal and Statistical Analysis of the Election Deadlock and the Ensuing Litigation, 2000 S Ct Rev 1, where I also discuss at greater length the issues discussed in this Essay. 2 See id at 2.

2 The University of Chicago Law Review [68:719 so that a hand recount might show that Gore had really won. A hand recount might indeed have shown this, but it would not follow that Gore had really won. The hand recount might be as unreliable as the machine count that it was intended to correct. While machines can be poorly designed, defectively manufactured, and inadequately maintained, and as a result make many errors, human counters can be fatigued, biased, or simply unable to infer, with any approach to certainty, what the voter's intent was from a ballot that the machine refused to count; so they can make many errors too. The Democrats asked for hand recounts in only four counties (Broward, Palm Beach, Miami-Dade, and Volusia), in all of which Gore had prevailed in the machine count and the canvassing boards were dominated by Democrats. A recount in these circumstances was highly likely to produce more votes for Gore than for Bush; indeed, all Gore had to do was to pick up the same percentage of recovered votes as he had won in the machine count to increase his lead in these counties.' By the same token, a recount in counties won by Bush would have been likely to increase Bush's lead in those counties, though not necessarily by enough to offset Gore's gain in the four counties. Even without regard to the possible effects of recounts in other counties, Gore would have been likely to overcome Bush's statewide lead only if the votes in the four counties that Gore sought recounts in had been recounted in accordance with the criteria used in Broward County. Gore's net gain of 582 votes in the Broward County recount represented 0.15 percent of his total votes there; that is, the recount gave him a net addition of 15 votes for every 10,000 votes that he had obtained from the original tabulation. If his votes in Palm Beach and Miami-Dade counties are multiplied by the same percentage and the product added to his net gain in Broward, his aggregate net gain is 1,480 votes and overcomes Bush's 930-vote lead.' But the 1,480 figure is suspect, and not only because Gore's margin in the other two counties was smaller than his margin in Broward, and so he could not have expected to pick up as large a share of the recovered votes; and also not only because of the domination of the canvassing boards by De- 3 To illustrate, suppose that in some precinct Gore had 7,000 votes counted by the machine and Bush 3,000, and 300 had not been counted, so that Gore led Bush by 4,000 votes. If now the 300 were counted and they split in the same proportion as the votes that had been counted previously, Gore would get 210 more votes and Bush 90 more, increasing Gore's lead by 120 votes. 4 Volusia County completed its hand recount before the November 14 statutory deadline and therefore the 98-vote net gain that it produced for Gore was included in calculating Bush's 930-vote official margin. As we will see, the Volusia recount probably produced an excessive net gain for Gore.

3 2001] Prolegomenon to an Assessment mocrats, who would be likely to give the edge to Gore in close cases. Democratic domination of the boards also made them likely to use criteria designed to maximize the number of recovered votes, since, as I have explained, the more votes recovered in counties Gore had carried, the more Bush's lead would erode. But when the criteria designed to maximize the number of recovered votes, which is to say the criteria used by Broward County's canvassing boards, are projected to the other counties to give Gore an estimated net gain of 1,480 votes, these projections are bound to be unreliable. This point is essential, and must be explained. The counties in question used the punchcard voting method. A card is placed on a tray, and the voter votes by punching a hole next to the candidate's name. The piece of the card thus dislodged, the "chad," falls to the bottom of the tray. The card is then removed and placed in a machine that counts votes by beaming light through the holes. If the chad is not punched through, the light will be blocked and the vote not registered. A chad that though punched remains dangling from the ballot by one or two corners, with the result that the vote was not counted by the tabulating machine, may be pretty good evidence of an intent to vote for the candidate whose chad was punched, provided the voter did not also punch the chad of another candidate for the same office (a significant qualification, however, as we will see). But inferring a voter's intentions from a merely dimpled chad, or a chad only one of whose corners has been separated from the ballot, is chancy. Nevertheless, Broward County's canvassing board apparently counted all such chads in undervotes (ballots that the machines had recorded as containing no vote for a presidential candidate) as valid votes. It did this even though a faint dimple might be created by the handling of the ballot or by its being repeatedly passed through the vote-counting machines; even though the voter may have started to vote for the candidate but then changed his mind, perhaps realizing he had made a mistake; even though there were many undecided voters in the 2000 presidential election, some of whom may have gone into the voting booth still undecided and, in the end, "decided" they could not make up their minds which presidential candidate to vote for; and even though some voters (probably more than the undecideds) undoubtedly misunderstood the ballot instructions. Those instructions were clear, but only if one could read. Voters voting on the basis of their recollection of oral instructions received from party activists would be bound often to make a mistake and spoil their ballot. In addition, the punchcard method of voting used in these counties requires a minimum of manual dexterity, and some voters lack even that.

4 The University of Chicago Law Review [68:719 Because the undercounted votes were only a small fraction of the total number of votes cast, it would not be surprising if a large fraction of them had been cast by undecided, confused, clumsy, inexperienced, or illiterate voters. This inference is especially compelling in the case of ballots in which the voter punched through the chads of all but the presidential candidates, indicating that the voting machine was not defective. This was the ground on which the Palm Beach County Canvassing Board decided eventually to exclude such ballots while counting those that had several, though apparently as few as three, dimpled chads. It was also why the Miami-Dade County Board could not, had it decided to count dimpled ballots, reasonably have confined the recount to the 10,750 undervotes in that county. Some of the ballots counted for the presidential candidate whose chad had been punched through may have contained a hanging chad in the other presidential candidate's hole. A hand recount would suggest that the voter had voted for two presidential candidates, voiding the ballot. Yet a ballot in which the voter punched through both presidential chads but left one dangling is as good evidence that the voter tried to vote for both candidates as a dangling chad in an undervote is evidence that the voter tried to vote for that candidate. It would not take a large percentage of such ballots, out of the more than 600,000 ballots cast in Miami-Dade County, to offset irregular ballots among the 10,750 undervoted ballots. The hand-recount criterion most favorable to the Democrats yet at least minimally objective was Palm Beach's three-dimples rule. A more conservative method would have counted dimples only in ballots in which no chads had been punched through, a pattern particularly suggestive of the voter's having tried unsuccessfully to punch through and perhaps been thwarted by a defect in the voting machine (for example, chad buildup). Palm Beach's method produced either 176 or 215 extra votes for Gore; the Florida Supreme Court declined to decide which number was correct, leaving the matter to further proceedings that were interrupted before the correct number could be determined. Assume, favorably to Gore, that the higher one was correct. It was still only 0.08 percent of the votes (8 in 10,000) that the machine had counted for Gore in Palm Beach County. Had the Broward and Miami-Dade Canvassing Boards used the Palm Beach method and produced the same percentage of additional votes for Gore, he would not have overtaken Bush's lead. His total gain in all three counties would have been only 788 votes (215 in Palm Beach as

5 2001] Prolegomenon to an Assessment mentioned, 263 in Miami-Dade, and 310 in Broward) The 168 additional votes that Gore netted in Miami-Dade from the first 20 percent or so of the recounted precincts before the recount was interrupted are a meaningless figure, because these precincts are far more heavily Democratic than the county as a whole. Of the disputed votes awarded to either Gore or Bush that yielded Gore's net gain of (at most) 215 votes in Palm Beach County, 61 percent went to Gore and 39 percent to Bush, compared to a 62 percent/38 percent split of the total machine-counted Palm Beach vote. This suggests that Gore would probably not have received more than 50 percent of the undervotes in a statewide hand count, since that was his percentage of the statewide machine count. It also underscores the meaninglessness of the 168-vote gain for him from the partial recount in heavily Democratic precincts in Miami-Dade County. He received 70 percent of the additional votes recorded by the recount, though his margin in the county as a whole was only 53 percent. An alternative method of estimating how the undervotes might have split between the two candidates is to compare Gore's vote gain in Palm Beach County with the number of disputed ballots in that county, 14,500. His net vote gain of 215 from that batch of ballots was only 1.5 percent of the votes that he got from that batch in the machine count. The same percentage of the 10,750 undervotes in Miami- Dade County would have meant a net gain of only 161 votes, compared to 263 by my earlier method. And even the 161-vote estimate is inflated. Gore received a lower percentage of the total Gore-Bush vote in Miami-Dade County-53 percent, compared with 62 percent in Palm Beach County. If a 24 percent margin (62-38) would have yielded Gore 161 extra votes over Bush, a 6 percent margin (53-47) would have yielded him only 40. Finally, it might make sense to average the machine-count and hand-recount results, since if their errors are independent averaging the two results will cause many of the errors to cancel out. But averaging Bush's 930-vote machine lead with a smaller but still positive lead obviously would not produce a victory for Gore. Some voters in Palm Beach County who cast ballots that the machine counted were misled by the "butterfly" ballot (in which the candidates are listed on both sides of the ballot rather than all on one side) used in that county; and voted for Buchanan when they meant to vote for Gore. 6 This ballot, the brainchild of the Democratic supervi- 5 In addition, he would have had a net gain of only 78, not 98, votes in Volusia County, reducing his overall net gain from 788 to Bush was listed first on the left-hand side of the ballot and Gore second. Buchanan was

6 The University of Chicago Law Review [68:719 sor of elections for the county, was intended both to enable the candidates' names to be printed in large type, in consideration of the number of elderly voters in the county, and to place before the voter all the candidates for each office without need to turn the page. Another ballot design, while less confusing, would have disenfranchised an unknown number of voters who had poor eyesight or cast their vote before realizing that there were additional candidates on the next page of the ballot. Even if, as widely and I think correctly believed, the butterfly design was on balance a mistake, it was an irremediable one for purposes of the 2000 election-not only because there was no reliable method of determining within any reasonable deadline for selecting Florida's electors the true intent of these voters, but also because altering an election outcome on the basis of the confusing design of the ballot would open a Pandora's box of election challenges. Gore's best chance for overcoming Bush's lead, it turns out, as we will see in the next Part, would have been to recount overvotes as well as undervotes; for there is evidence of recoverable overvotes, in particular ballots in which the voter had both punched a candidate's chad and written in the name of the same candidate in the place for write-in votes. But Gore did not want overvotes recounted. In all likelihood, therefore, a fairly designed and administered hand recount would not have enabled Gore to overcome Bush's 930-vote lead. Let us now see what the Florida Supreme Court did to unsettle that lead. II. WHAT THE FLORIDA SUPREME COURT DID TO FLORIDA'S ELECTION STATUTE A. The Statute Florida's election statute requires counties to certify their vote totals within seven days of the election, which in 2000 meant by November 14,' except that overseas ballots are as a consequence of federal law to be counted up to the tenth day after the election and added to the seventh-day totals. 8 Up to that seventh day a candidate may "prolisted first on the right-hand side, between Bush and Gore rather than opposite to Bush. The candidates' chads were in the middle of the ballot, which meant that Buchanan's chad was between Bush's and Gore's. A voter who wanted to vote for Gore had to punch the third chad down; if he punched the second, which was almost level with the word "Democratic" above Gore's and Lieberman's names, he was voting for Buchanan. It would be an easy mistake to make, especially for an inexperienced voter or one with poor eyesight, though, as mentioned in the text below, the butterfly format enables larger type and in that respect helps people with poor eyesight. 7 Fla Stat Ann (1) (West 2000). 8 See Palm Beach County Canvassing Board v Harris, 772 S2d 1273, 1288 (Fla Dec 11,

7 2001] Prolegomenon to an Assessment test" the result of the election in a county as "being erroneous" and ''may... request... a manual recount" and the county canvassing board "may authorize" it. 9 This hand recount is of just a sample of precincts, but if it "indicates an error in the vote tabulation which could affect the outcome of the election," ' the board must take corrective action, which can include a hand recount of all the ballots cast in the county." Should that recount not be completed by the seventh day, its results "may be ignored" by the Florida Secretary of State.' 2 Once the Secretary of State has received the certified county totals and certified the election winner, the loser can "contest" the result by filing a lawsuit." 3 If he can prove that enough "legal votes" were rejected to have "change[d] or place[d] in doubt the result of the election," the court can "provide any relief appropriate under such circumstances."" A "damaged or defective" ballot is not to be declared invalid if it contains "a clear indication of the intent of the voter."' 5 None of the hand recounts sought by Gore, except the one in Volusia County," was complete by November 14, and the Secretary of State said she would refuse to consider them. She interpreted the statute to mean that unless there was evidence of fraud or statutory violations, or some disaster (a hurricane, for example) that had interrupted the recount, the seven-day deadline was firm.' 7 The Director of the Division of Elections in the Secretary of State's office interpreted, presumably with her concurrence, the statutory term "error in the vote tabulation" to mean a failure of the tabulating machine to count properly marked ballots, rather than the machine's failing to record a vote because the voter had failed to follow the instructions for casting a valid, machine-readable vote or to complain to a precinct worker if the voter could not follow the instructions because the voting machine was defective. 8 The instructions were clear. The many dimpled and 2000). 9 Fla Stat Ann (1), (1), (4)(a), (c) (West 2000). 10 Id (5). 11 Id (5)(c). 12 Id (1). The preceding Section ( (1)) says "shall be ignored," creating the only real inconsistency in the statute. I have no quarrel with the Florida Supreme Court's preferring "may," which was also the interpretation of the Secretary of State. 13 Id (1). 14 Id (3)(c), (e)(8). 15 Id (5). 16 See note Her written statement is quoted in Palm Beach County Canvassing Board v Harris, 772 S2d 1220, n 5 (Fla Nov 21,2000), vacd and remd as, Bush v Palm Beach County Canvassing Board, 121 S Ct 471 (2000) (per curiam). 18 Roberts's interpretation can be found in the Joint Appendix to the Respondent's Supplemental Brief, Bush v Palm Beach County Canvassing Board, No , *52-58 (filed Nov 30,

8 The University of Chicago Law Review [68:719 dangling chads were the result of voters' either failing to follow the instructions or, if the voting machine itself was defective, so that the instructions could not be followed, failing to seek the assistance of an election official -a failure that was also a form of voter error. If voter error was not a valid basis for a hand recount, there was no possible justification for extending the statutory deadline for submission of a county's votes in order to permit an effort to recover votes from ballots rejected because of voter error. The only reason the county canvassing boards needed extra time was to complete the laborious hand recounts necessary to recover votes from ballots that the voters spoiled, a process of interpretation rather than of mere inspection. When on November 21 the Florida Supreme Court reversed the Secretary of State and extended the November 14 deadline for protest recounts to November 26,'9 the consequence was to postpone till then the certification of the election results and hence the commencement of the contest proceeding, thus squeezing the time for completing such a proceeding so tightly as to make completion by any realistic deadline chancy. This implication of the court's ruling suggests that the Secretary of State's statutory interpretation was the correct one after all. As does the statutory text: "error in vote tabulation" does not sound like an error by the voter; it sounds like an error by the mechanical or human tabulator. The machinery for counting punchcard ballots, which are the form of ballot used in 40 percent of Florida's counties, containing 63 percent of the state's population, was not designed to tabulate dimpled or otherwise unpunched-through ballots; so how could its failure to count such ballots be thought an error in tabulation? And because that "failure" was built into the design of the machine, to deem it the result of an error in vote tabulation would make hand recounts mandatory in all close elections in the many counties that use punchcard voting machines. The legislature was unlikely to have intended this when it passed the statute, especially given the vagaries of hand recounting of spoiled ballots ) (available on Westlaw at 2000 WL ) (including letters from Roberts regarding the meaning of error in tabulation). 19 Palm Beach County Canvassing Board v Harris, 772 S2d at In contrast, the absence of a postmark on some absentee ballots of military personnel, a defect the Secretary of State did not think invalidated those ballots, was akin to a tabulating error. The voter does not affix the postmark. The overseas voters had followed instructions to the letter and had thus done all they could do to cast a legal vote; the mistake was in the transmission machinery. The Democrats, after being accused of being anti-military, decided not to make an issue of the absence of postmarks on military ballots. Another irregularity in the processing of absentee ballots, however, was challenged. Republican campaign workers had been permitted to enter the election offices in two counties to affix voter identification numbers to absentee ballots that

9 20011 Prolegomenon to an Assessment The statute does not specify the circumstances if any in which the Secretary of State is required or even permitted to include in her certification of the election results the results of a recount not completed by the statutory deadline. But the statute authorizes the Secretary of State to interpret the statute," implying that her interpretation if reasonable is conclusive. The interpretation by the director of her division of elections, the interpretation that voter error is not a ground for extending the deadline, was reasonable and should therefore have been conclusive on the Florida Supreme Court. Indeed, it was the natural and sensible interpretation of "error in the vote tabulation," which, to repeat, is the only basis in the statute for a complete hand recount of a county's votes." Against this conclusion, Michael McConnell, without addressing the question of the Secretary of State's interpretive authority, argues that the election statute does not limit the reasons for conducting a manual recount." But this is incorrect. A full countywide manual recount (as distinct from the initial sample recount) is permitted only "if the [sample] manual recount indicates an error in the vote tabulation which could affect the outcome of the election." ' McConnell also cites a provision of the election law which states that if during the manual recount the counters are "unable to determine a voter's intent in casting a ballot, the ballot shall be presented to the county canvassing board for it to determine the voter's intent.' 5 But as McConnell himself notes, the Florida Supreme Court never cited this provision; and I cannot see its bearing on the meaning of "error in the vote tabulation." If because of such an error all the ballots have to be inspected by hand, some of them are bound to be spoiled ballots; and if it is unclear whether or not a ballot is spoiled, the question is properly referred to the canvassing board. That is all the provision in question means. The statute also provides that the Secretary of State may ignore recount results received after the seventh day following the election. This implies that she has discretion to ignore results from recounts not were missing them. The Florida courts rebuffed the challenge on the ground that the voters' intentions were clear. See Jacobs v Seminole County Canvassing Board, 773 S2d 519, 2000 Fla LEXIS 2404, *5-9 (Dec 23) (per curiam). The numbers were affixed before the votes were counted, and so there was no question of an error in the tabulation of the votes. 21 Fla Stat Ann (1) (West 2000). 22 Id (5). 23 Michael W. McConnell, Two-and-a-Half Cheers for Bush v Gore, 68 U Chi L Rev 657, (2001). 24 See Fla Stat Ann (5), (5)(c). 25 Id (7)(b); McConnell, 68 U Chi L Rev at 664 (cited in note 23). 26 Fla Stat Ann (1).

10 The University of Chicago Law Review [68:719 barred by the statute (for example, a hand recount conducted because of an error in the vote tabulation that might have affected the outcome of the election) as well as being compelled to ignore those that are barred." So even if she erred in thinking her hands tied by the statute, and therefore failed to exercise her discretion to ignore or not to ignore late recount results designed to correct voter error, the court should have directed her to exercise her discretion; that is the remedy for a failure to exercise discretion.2 It would have been well within her discretion to ignore the late recount results -as the expense, delays, near riots, litigation avalanche, and general turmoil that attended the Florida Supreme Court's refusal to uphold her decision demonstrate. B. What the Florida Supreme Court Did to the Statute The Florida Supreme Court reversed Judge Lewis and extended the statutory deadline to November 26, at the same time holding that the recount should include spoiled ballots in which the voter's intent was discernible. The justices thought it okay to strong-arm the statute because they thought it internally inconsistent inasmuch as it allowed a recount to be sought right up to the seventh day after the election even though a recount requested on the last day could not be completed by the end of that day, the deadline for submission by the counties of their vote totals. 29 There is no inconsistency. If the recount is not requested promptly after the election and so cannot be completed by the seventh day, the losing candidate has mainly himself to blame for not having acted faster; and the only consequence is that he must fall back on his remedy of filing a contest proceeding. The Florida Supreme Court in its second opinion, that of December 8, reversing the 27 As Judge Lewis explained in upholding her refusal to include late recount results, "Florida law grants to the Secretary [of State], as the Chief Elections Officer, broad discretionary authority to accept or reject late filed returns." McDermott v Harris, 2000 WL , *1 (Fla Cir Ct Nov 17), revd as Palm Beach County Canvassing Board v Harris, 772 S2d 1220 (Fla Nov 21, 2000), vacd and remd as Bush v Palm Beach County Canvassing Board, 121 S Ct See, for example, Chathas v Local 134 IBEW, 233 F3d 508, 514 (7th Cir 2000) (remanding the case to enable lower court judge to make the required discretionary judgment); Campanella v Commerce Exchange Bank, 137 F3d 885, (6th Cir 1998) (remanding because the "district court appears not to have recognized that it had supplemental jurisdiction, and thus failed to actually undertake an exercise of discretion"); Channell v Citicorp National Services, Inc, 89 F3d 379,387 (7th Cir 1996). As the court said in Channell, "Because he held that [28 USC] 1367(a) did not authorize the exercise of supplemental jurisdiction, [the district judge] did not exercise the discretion 1367(c) confers. It belongs to him rather than to us, so we remand for its exercise." Channell, 89 F3d at Palm Beach County Canvassing Board v Harris, 772 S2d 1220, 1235 (Fla Nov 21,2000), vacd and remd as, Bush v Palm Beach County Canvassing Board, 121 S Ct 471 (2000) (per curiam).

11 2001] Prolegomenon to an Assessment dismissal of the contest proceeding, ruled in effect (though erroneously, as I shall argue) that the certification has no presumptive validity in such a proceeding. This implies that the disappointed candidate loses very little by being remitted to his contest remedy. The more important point is that seven days is plenty of time to correct an error in vote tabulation, even when a hand recount is necessary, since the voter's intention in a ballot that has been filled out in accordance with the instructions can be determined by a simple inspection of the ballot; debatable interpretation is not required. The Secretary of State was entitled in the exercise of her discretion in the interpretation and application of the statute to conclude that wanting to recover votes from ballots spoiled by the voter was not a proper reason for an extension of the statutory deadline-especially in a presidential election, where delay in certifying the results of the election could well cause chaos." It is precisely in adapting the statute to the exigencies of a presidential election that the Secretary of State might have been expected to be given a freer rein in statutory interpretation and application. Her decision not to delay the certification of the winner of the presidential election deserved considerable deference; it received none. The Florida Supreme Court made no effort to conceal the fact that in interpreting the statute differently from the Secretary of State it was appealing to a higher law than the statute. It derided "sacred, unyielding adherence to statutory scripture" 32 and "hyper-technical reliance upon statutory provisions" 3 and said that "the abiding principle governing all election law in Florida" was to be found in the statement 30 Gore v Harris, 772 S2d 1243, 1260 (Fla Dec 8,2000) (holding that votes hand counted in the contest proceeding "must be included in the certified vote totals"), revd and remd as, Bush v Gore, 121 S Ct 525 (2000) (per curiam). 31 See Posner, 2000 S Ct Rev at 27 (cited in note 1). 32 Palm Beach County Canvassing Board v Harris, 772 S2d at 1228, quoting Boardman v Esteva, 323 S2d 259,263 (Fla 1975). The court did not, however, quote the following passage from Boardman: "the results of elections are to be efficiently, honestly and promptly ascertained by election officials to whom some latitude of judgment is accorded, and... courts are to overturn such determinations only for compelling reasons when there are clear, substantial departures from essential requirements of law." 323 S2d at n 5 (quoting the trial court with approval). The question in Boardman -and in another decision from which the Florida Supreme Court in Palm Beach County Canvassing Board v Harris quoted sonorous right-to-vote language, Beckstrom v Volusia County Canvassing Board, 707 S2d 720, 725 (Fla 1998) -was whether substantial compliance with the rules governing absentee ballots was sufficient to allow a vote to be counted. The court held that it was, thus presaging its decision regarding irregularities in the absentee ballots challenged in Seminole and Martin Counties in the 2000 election. See note 20. Compare Chappell v Martinez, 536 S2d 1007, (Fla 1988), and Carpenter v Barber, 198 S 49 (Fla 1940). 33 Palm Beach County Canvassing Board v Harris, 772 S2d at 1227.

12 The University of Chicago Law Review [68:719 in Florida's constitution that "all political power is inherent in the people." The court was using the Florida constitution, or perhaps some principle of natural law, to trim the statute. "[T]he will of the people is the paramount consideration... This fundamental principle, and our traditional rules of statutory construction, guide our decision today." 3 Armed with a principle as vague and all-encompassing as "people power," a court can do anything with an election law in the name of interpretation. The court's later explanation that it had been using the ordinary principles of statutory interpretation after all, and specifically that the "plain meaning" of the statutory term "error in the vote tabulation" included an error resulting from a voter's mistake that made his ballot unreadable by the machine, was lame. The Secretary of State's interpretation was the plain meaning of that term. 7 After the Secretary of State on November 26 (the court-imposed extended deadline for certification following protest) certified Bush as the winner, albeit with a diminished lead (Broward having completed its recount by then), Gore brought suit against the Democratic canvassing boards of Palm Beach and Miami-Dade, 8 contesting the election results in those two counties. On December 4, after a two-day trial, Florida Circuit Judge Sauls dismissed Gore's suit.3 Gore appealed and four days later the Florida Supreme Court reversed in another indefensible opinion.- Judge Sauls had interpreted Florida's election statute as establishing the contest as a judicial proceeding to review administrative action." The purpose of the contest trial was thus to 34 Id at 1230, quoting Fla Const Art I, Palm Beach County Canvassing Board v Harris, 772 S2d at 1228 (emphasis added). 36 Palm Beach County Canvassing Board v Harris, 772 S2d 1273, (Fla Dec 11, 2000) (per curiam). 37 David A. Strauss, in his contribution to this symposium, Bush v Gore: What Were They Thinking?, 68 U Chi L Rev 737,752 (2001), argues that the Florida Supreme Court's interpretations of other provisions of the election statute, such as "legal vote," were not "inconsistent with the plain language of the contest statute" and therefore that "there is a plain language defense for the Florida Supreme Court's action." This suggests a misunderstanding of the "plain meaning" interpretive principle. The principle is that if statutory language is plain, that is, clear as a linguistic matter (regardless of the real world context of application, which might expose an ambiguity), the court should follow it, unless the result is totally absurd. All Strauss means is that some of the statutory language is not plain. But "error in the vote tabulation" is plain, and it is merely bizarre for the Florida Supreme Court to have claimed that its interpretation, which violated the plain meaning of the term, was the plain meaning. 38 Also Nassau County, but that part of Gore's case went nowhere, and I will ignore it. 39 Gore v Harris, 2000 WL , *5 (Fla Cir Ct Dec 3), revd and remd as Gore v Harris, 772 S2d 1243, revd as Bush v Gore, 121 S Ct Gore v Harris, 772 S2d The vote this time was 4-3 (the November 21 decision had been unanimous), with powerful dissents. 41 Gore v Harris, 2000 WL at *4.

13 20011 Prolegomenon to an Assessment determine whether the canvassing boards, the organs charged with tabulating election results, had abused their discretion in failing to conduct a hand recount in a certain way (that is, the Broward way, sought by Gore), or at all (the Miami-Dade County Canvassing Board had begun a hand recount, then changed its mind and stopped). 2 As there was no reason to believe that the result of the statewide election would change in favor of Gore if reasonable recounting procedures were followed, 4 3 of which the most favorable method to Gore that could be considered reasonable was Palm Beach's three-dimples method, Judge Sauls found no abuse of discretion and so refused to order a further recount. By reversing him and holding that the decision of a canvassing board is entitled to no deference in a contest proceeding, the Florida Supreme Court made the protest a meaningless preliminary to the contest and expanded, without any basis in the statute, the power of the courts relative to that of the officials (the members of the canvassing boards and the Secretary of State) to whom the legislature had actually confided the conduct and supervision of elections, including election recounts. And while the statute as we have seen limits the canvassing boards to correcting errors in the vote tabulation," the December 8 opinion authorizes (even requires) judges in contest cases to conduct recounts intended to rectify voter errors as well. 4' The court, even though it lacks staff and experience for counting and interpreting ballots (especially thousands or tens of thousands or millions of ballots), becomes the primary tabulator, rather than the election officials. That is upside down.46 It is true that the election statute does not confine contests to situations in which there has been an error in the vote tabulation. The grounds include fraud or other misconduct by an election official, bribery, the counting of illegal votes, and the ineligibility of a candidate. 4 7 The only ground available to Gore, however, was the failure to count "legal votes." And in context that was merely a complaint about the canvassing boards' handling of spoiled ballots. The principles of administrative law required the contest court, as Judge Sauls ruled, to defer to the canvassing boards as the experts in counting votes, and thus to uphold their decisions unless unreasonable. Given the problems of interpreting spoiled ballots, the decisions of the Palm Beach 42 Id. 43 Id at *3. 4 Fla Stat Ann (5). 45 Gore v Harris, 772 S2d at The Chief Justice forcefully argued this point in his dissenting opinion. Id at (Wells dissenting). 47 See Fla Stat Ann (3).

14 The University of Chicago Law Review [68:719 and Miami-Dade boards were not unreasonable, as Judge Sauls correctly concluded. As well as upsetting the balance between court and agency, the Florida Supreme Court set the threshold for relief in a contest proceeding at an implausibly low level. No human or machine fault in the conduct of the election, and no external circumstances (such as a natural disaster) that might interfere with the conduct of the election, had to be shown. It was enough that the election had been close and that a hand recount using unspecified criteria might recover enough undervotes to change the outcome. ' Successful contests, in the sense of contests eventuating in judicial orders for selective or comprehensive hand recounts, would become the norm in close elections. On November 21, the Florida Supreme Court had extinguished the Secretary of State's discretion. On December 8, it extinguished the canvassing boards' discretion. The justices said in effect: if the election is close and we think there were a lot of voter errors, we have carte blanche to order any mode of recount that strikes us as likely to recover a substantial number of the rejected votes. The Florida election statute could provide that electors are to be picked by the state's supreme court after it knows (and maybe does not like) the result of the election, using a standard of the voter's unclear intent and the principles of natural law, even when there is no reason to suppose that an infallible hand recount would reverse the result of the election. But the legislature did not say anything like that. The court ordered that Gore's 215-vote net gain in the Palm Beach recount (or 176-the court left it to the trial court to decide which number was correct), plus his 168-vote net gain from the partial recount in Miami-Dade County, be added to Gore's certified total. 9 The court thus was changing that total after the deadline (November 26) that it itself had set for determining the certified vote totals after the completion of the protest recounts. The court ordered a hand recount of all the remaining undervotes not only in Miami-Dade but throughout the state, 0 which have been estimated at sixty thousand. The trial before Judge Sauls had made crystal clear, if it was not already, that the hand recounts were neither uniform" 1 nor reliable. There were Broward rules, which favored Gore unduly, as we have 48 Gore v Harris, 772 S2d at Id at 1248, Id at See, for example, Trial Transcript, Gore v Harris, No , * (Fla Cir Ct Dec 2, 2000) (available on Westlaw at 2000 WL ) (testimony of Judge Charles Burton, chairman of the Palm Beach canvassing board).

15 2001] Prolegomenon to an Assessment seen, and a medley of different Palm Beach rules, the three-dimples rule having emerged after the recount had begun (earlier iterations had been a "sunshine rule"-light must be visible through the chad hole-and a "no dimple" rule, the rule followed in Palm Beach county in previous recounts 52 ). No one knew what standard the Miami-Dade counters had used before they abandoned the recount. Yet while acknowledging that "practical difficulties may well end up controlling the outcome of the election" 53 (that is, may terminate the contest proceeding before its completion), the Florida Supreme Court gave Gore the votes he had gained in Miami-Dade County before the recount was interrupted, even though the precincts counted were unrepresentative and time might be called on the complete recount. The court refused to prescribe a uniform standard, and thus was ordering a recount that could not be expected to be accurate, or even concluded. Yet if Gore was ahead in the recount when time was called, the court's opinion implied (probably inadvertently) that he would be declared the winner even if the disputed ballots in precincts likely to favor Bush had yet to be recounted. For the court had given Gore the votes he had received in Miami-Dade's partial recount even though the full recount might never be completed. Critics of the U.S. Supreme Court's intervention in the election litigation blame that Court for the Florida Supreme Court's failure to set a uniform standard. The argument is that by doing so the Florida court would have stoked the fires of Bush's argument based on Article II of the U.S. Constitution that the court was revising rather than interpreting the election statute. But what is more likely to have stopped the court (since filling a gap in a statute is not rewriting the statute) is that a uniform standard based on Broward procedures would have been completely untenable, yet a uniform standard that was inconsistent with those procedures would have made the inclusion of the Broward recount result in Gore's certified vote total nonsensical-it would amount to crediting Gore with votes recovered by a procedure that the court itself was rejecting as unreliable. There is more wrong with the December 8 opinion. In ordering that only undervotes be recounted, the court was ignoring the fact that Gore's gains in Broward, Palm Beach, Miami-Dade, and Volusia counties may have included overvotes. An overvote is less likely to be recovered by a hand recount than an undervote is. But if the chad for one candidate is cleanly punched through and the chad for his rival 52 See id. 53 Gore v Harris, 772 S2d at n 21.

16 The University of Chicago Law Review [68:719 slightly dislodged because the voter started to vote for the rival and then realized he was making a mistake, the machine might read the second dislodgement as a vote and void the ballot. And if the voter both punched a candidate's chad and wrote the candidate's name in the space provided for write-in votes, the machine would automatically reject the ballot, even though the voter's intention was plain.' There were 110,000 overvotes statewide, so given the closeness of the election the refusal to order them recounted could not easily be justified, except for the shortness of time. More to the point, if Gore was given votes recovered from overvotes in the four counties' recounts, those ballots had to be recounted, because some of them may well have been true overvotes, where, as I explained earlier, the voter had punched through the chads for two presidential candidates but one of the chads had been left hanging and so the machine, which cannot be relied on to count hanging chads as votes, had failed to reject the ballot. Despite the shortness of time, the Florida Supreme Court assumed on December 8 that a responsible recount might be completed and the state's electoral votes certified by December 12; electoral votes certified after that date could be challenged in Congress when the votes were counted in January." There was no way in which sixty thousand votes could be recounted by the twelfth yet allow time for the contestants' lawyers to challenge, and a judge to review, the decisions made by the counters on particular ballots, especially when the counters would be using different criteria for what constituted a "legal vote." Either the court meant to condone a bobtailed procedure or it was expecting the recount to fizzle and did not want to be blamed. Granted, the Florida statute is vague when it comes to relief. Remember that the court in a contest proceeding that finds that enough "legal votes" were rejected to "change or place in doubt the result of the election" can "provide any relief appropriate under such circumstances. 56 "Appropriate" is not defined; it has been left to the courts to work out on a case-by-case basis. But even a term as vague as "appropriate" does not give a court carte blanche. No reasonable person could consider the relief ordered by the Florida Supreme Court on December 8 appropriate. Once again the court had misinterpreted it. 54 Gore's own people thought that the second type of error had been made in Duval County and had cost Gore a significant number of votes. Richard T. Cooper, A Different Florida Vote, LA Times Al (Dec 24, 2000). According to Cooper, the Gore team did not discover the overvote problem in Duval County in time to request a recount there USC 5 (1994). 56 Fla Stat Ann (3)(c), (e)(8).

17 20011 Prolegomenon to an Assessment CONCLUSION It is natural to conclude that if the critics are right and the U.S. Supreme Court lacked adequate grounds for intervening in the election litigation and therefore should not have halted the recount ordered by the Florida Supreme Court on December 8, an injustice was done. But if my analysis is correct, the conclusion is erroneous, if by "injustice" one means that Gore would have been the legal winner had the recount continued. Even with Bush's lead carved down to as few as 150 votes by the Florida Supreme Court's decision of December 8, it is unclear whether Gore would have prevailed in the recount ordered by that court. It is true that he needed to pick up only a few more than that number of votes from the nine thousand undervotes not yet recounted in Miami-Dade County to pull ahead, assuming the other fifty thousand or so undervotes statewide that had not yet been recounted would have split evenly. We do not know what rule the Miami-Dade canvassing board would have used had it resumed recounting. But if it would have used Palm Beach rules on all 10,750 disputed ballots in Miami-Dade County, then by my earlier analysis it probably would have given Gore only 263 additional votes-fewer than 100 more than the 168 vote gain that the Florida Supreme Court had already given Gore in Miami-Dade. Had Gore gained only 100 more net votes, Bush would still have won the state, albeit by as few as 50 votes. Gore might have prevailed in the recount-miami-dade's canvassing board might have used Broward rules, with results similar to what they produced in Broward County-but only by virtue of the Florida Supreme Court's having violated state law. The decisions of November 21 and December 8 made a hash of that law. Abstention by the federal courts would not have erased the fact that the Florida Supreme Court had erred grievously in interpreting the Florida election law. It should not have extended the deadline for hand recounting that had been fixed by the Secretary of State, or interpreted "error in the vote tabulation" to include a voter's error in voting, or reversed Judge Sauls's dismissal of the contest proceeding, or extinguished the discretionary authority of the state and local election officials, or authorized relief in a contest proceeding on the basis merely that the election was close and there were a number of undervotes, or credited Gore with the Broward and the partial Miami-Dade recount results, or ordered a statewide recount of undervotes but not overvotes." 7 In all these re- 57 Gore might well have benefited from a hand recount of overvotes statewide, see note 54, but he did not seek such a recount. More to the point, voter error is not a basis under Florida law for recounting.

18 736 The University of Chicago Law Review [68:719 spects it was deforming Florida's election law. There was no legal basis for compelling the Secretary of State to accept late hand recounts; Bush really did win by 930 votes. Had Gore been declared the winner on the basis of the recount ordered by the Florida court on December 8, he would have owed his victory to legal error, whether or not it was a legal error that the U.S. Supreme Court should have corrected. The result of the Supreme Court's intervention was, therefore, at the least, rough justice; whether it was legal justice is the question I have left for another occasion.